1. Leveraging Process Data Historians: Data Analytics for Mechanical and Process Monitoring
Presented by:
John Antanies
Principal and Founder
Envoy Development, LLC

2. Data Historians
Data historians capture data from DCS and PLC systems, saving it for periods of years.
How many in the audience have a data historian such as PI, IP.21, ParcView, etc?

3. Reliability
Reliability suffers when maintenance issues that should have been identified are not
Reliability suffers when process issues that should have been identified are not
Examples:
Motors or bearings fail
Temperatures, amps, and vibration monitors typically increase before the failure
Flows suddenly become insufficient
Valves start to slowly open or close and eventually cause issues
Mixers, pumps and blowers fail but don’t go undetected because the motor is running
Belts or couplings on pumps/agitators/blowers shear but don’t set off an alarm

4. Most data historian users are still using them the same way they did in the early 1990s:
Looking at hundreds of trends
Dumping data into Excel
The result?
You miss things.
What if you could identify variables that are out of limits instantly, without resorting to trends and spreadsheets?
Think about this…

5. Operational Efficiency Essentials
Reliability: What issues need focus now?
Metrics: What drives KPIs?
Sustainability:
How do we preserve benefits?
Benchmarking: Are we operating within limits?

6. Data Collection Considerations
Data must be collected continuously
Data must be aggregated only when the process is running
Further processing must be done by grade/product and/or production rate
Limits can then be calculated:
Defined moving time frame (3 or 6 months)
Fixed time period (period where the process was understood to have ran well)
Best in class query (days or product runs that meet defined criteria)

7. Data Collection Example of data collection topology Server Side
dbEnvoy
Microsoft SQL Database
Envoy.DC
Data Collection Service
PI System
Custom Process
Database
AF SDK or PI SDK
Other Implementations
Hourly Aggregations
Daily Aggregations
Limits and Targets calculations
Per Work Shift Aggregations
Support Backfilling

8. When did costs change?
Defining when a Key Performance Indicator (KPI) changed is not as easy as one might think…
Or here?
Here?
Or somewhere else?

9. A CUSUM Chart Tells Us…
Costs actually changed twice, but most significantly on Jul 23.
When the line is going down, the variable is below average; the steeper the slope, the more negative the value.
When the line is going up, the variable is above average.
Points where the line goes from a negative to a positive slope (or vice-versa) are called inflection points.
If we calculate and index the CUSUM inflection points for historian variables, we can identify what else changed when costs changed.

10. Case Study: Web Breaks
We detected web breaks increased substantially on May 18
Remember, when the CUSUM line has a negative slope, the value is below the long term average; for web breaks, low is good.
On May 18 the line suddenly moved positive.
What changed?
CUSUM chart on left shows that breaks increased sharply after May 18

11. Statistical Analysis With CUSUM Filtering
If we look at tags that were different in a statistically different manner (high web breaks vs low) we have over 700 variables.
When we filter by CUSUM inflection we have less than 40 variables
Redacted

12. Web Breaks Finding
This customer removed an additive to try and save money
Once we identified this, they put it back on and web breaks declined

13. Case Study: Chemical Mixer
CLO2 MIXER CURRENT: 999-II-999
Two Sigma Limit : to 81.29
Current Value : 35.85
D/C Mixer Residual: BPRESDL.L
Two Sigma Limit : -0.2 to 1.09
Current Value : 2.81
The key takeaway from this case study:
If you don’t recognize what is going on, don’t be afraid to contact the equipment manufacturer.
A keyway corroded away and caused costs to increase $4,140/day
Customer Comment: Good find on the mixer, I spoke with GL&V this morning and they are pretty sure the mixer came off the shaft, we  have a spare in house and plan on changing it out this Wednesday
Costs skyrocketed when these two events occurred

14. Case Study: Sheet Caliper
We detected that caliper increased
There are two distinct periods of high caliper
What caused this process deviation?

15. Caliper Findings
A roll in a stack had been taken out of service; since it was behind a door, it went unnoticed

16. Questions?

17. Joh
Contact Information
John Antanies
Principal and Founder
Envoy Development, LLC
(office)
(mobile)